No network connection

[Jos den Bekker]

Yesterday installed Qubes 2nd release (Beta 3, I think), but no network connection. Spent hours trying to fix it, but to no avail. Network card is recognized. I hace a cable connection, no WiFi. I have a Dell Studio XPS 8100, Intel Core i3 CPU, 4 GB RAM, Broadcom Netlink Gigabit Ethernet card (Ethernet 802.3), 1 TB harddrive. I used option "Test and install" at installation, no error message. I have installed and used two other Linux flavors before (Ubuntu and Mint) and none of them had any trouble connecting me to the internet immediately (updates downloaded during install). Maybe installation went wrong, because both at restart and shutdown Qubes makes my computer hang. What's the best I can do? Download the iso again and do a complete reinstall? Wait for the next release? Buy another machine? ;-)

[Marek Marczykowski-Górecki]

Open netvm terminal (menu->Applications->ServiceVM: netvm->netvm: Terminal)
and check the following things:
1. Does 'lspci' list your network adapter?
2. Do you have interface detected (does 'ifconfig -a' contains en* device)?
3. Does kernel messages ('dmesg') contains some errors regarding network
device initialization?
4. Check if linux-firmware package is installed (rpm -q linux-firmware).

--
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?

[Jos den Bekker]

Op vrijdag 5 september 2014 01:44:21 UTC+2 schreef Marek Marczykowski-Górecki:

Open netvm terminal (menu->Applications->ServiceVM: netvm->netvm: Terminal)
and check the following things:
1. Does 'lspci' list your network adapter?
2. Do you have interface detected (does 'ifconfig -a' contains en* device)?
3. Does kernel messages ('dmesg') contains some errors regarding network
device initialization?
4. Check if linux-firmware package is installed (rpm -q linux-firmware).

1: Yes
2: No (see attachment)
3: Yes: [    8.918052] tg3 0000:00:00.0: Could not obtain valid ethernet address, aborting
           [    8.918261] tg3: probe of 0000:00:00.0 failed with error -22
           (see also attachment)
4: Not installed. How do I install it without internet connection?

Thanks in advance.

 

[Marek Marczykowski-Górecki]

You can download the package using other computer/system, then transfer it
with USB stick. URL to the package:
[user@netvm ~]$ yumdownloader --urls linux-firmware
(...)
http://mirror.karneval.cz/pub/linux/fedora/linux/updates/20/x86_64/linux-firmware-20140605-38.gita4f3bc03.fc20.noarch.rpm

[Jos den Bekker]

On monsay 8 September 2014 18:41:29 UTC+2  Marek Marczykowski-Górecki wrote:

You can download the package using other computer/system, then transfer it
with USB stick. URL to the package:
[user@netvm ~]$ yumdownloader --urls linux-firmware
(...)
http://mirror.karneval.cz/pub/linux/fedora/linux/updates/20/x86_64/linux-firmware-20140605-38.gita4f3bc03.fc20.noarch.rpm

 
Downloaded and installed it. I hope I did it the right way:

[user@netvm ~]$ sudo rpm -i path-to-rpm-file
[user@netvm ~]$ rpm -q linux-firmware
20140605-38.gita4f3bc03.fc20.noarch
[user@netvm ~]$

But still no internet connection. dmesg gives the same error message, ie:
[    8.918052] tg3 0000:00:00.0: Could not obtain valid ethernet address, aborting

[    8.918261] tg3: probe of 0000:00:00.0 failed with error -22

Strange, though, because, as I said before, I have never had internet connection problems with any Linux distribution on this machine. Should something be purged as well?

Thanks a lot so far and best regards.

Jos den Bekker

[Marek Marczykowski-Górecki]

On 08.09.2014 22:07, Jos den Bekker wrote:
>
>
> On monsay 8 September 2014 18:41:29 UTC+2 Marek Marczykowski-Górecki wrote:
>>
>>
>> You can download the package using other computer/system, then transfer it
>> with USB stick. URL to the package:
>> [user@netvm ~]$ yumdownloader --urls linux-firmware
>> (...)
>>
>> http://mirror.karneval.cz/pub/linux/fedora/linux/updates/20/x86_64/linux-firmware-20140605-38.gita4f3bc03.fc20.noarch.rpm
>>
>>
> Downloaded and installed it. I hope I did it the right way:
> [user@netvm ~]$ sudo rpm -i path-to-rpm-file
> [user@netvm ~]$ rpm -q linux-firmware
> 20140605-38.gita4f3bc03.fc20.noarch

Ok, so try to unload and load again tg3 module.

Generally, you should install that package in the template VM (fedora-20-x64)
to have that change persistent.

[Marek Marczykowski-Górecki]

On 09.09.2014 06:06, Fernando Parodi wrote:

> Hi Guys, I-m having Internet connection Problems:

>
> 1. Does 'lspci' list your network adapter?

> YES

>
> 2. Do you have interface detected (does 'ifconfig -a' contains en*

>>> device)? - YES

>
> 3. Does kernel messages ('dmesg') contains some errors regarding network

> device initialization? - Yes
>
> Detail:
>
> [ 7.692223] systemd-udevd[519]: renamed network interface eth0 to enp0s1
> [ 7.752864] ath5k 0000:00:00.0: enabling device (0000 -> 0002)
> [ 7.753206] ath5k 0000:00:00.0: Xen PCI mapped GSI17 to IRQ82
> [ 7.753583] ath5k 0000:00:00.0: registered as 'phy0'
> [ 7.776253] systemd-udevd[520]: renamed network interface eth1 to enp0s2
> [ 8.452629] cfg80211: Calling CRDA for country: CN
> [ 8.452672] ath5k: phy0: Atheros AR2413 chip found (MAC: 0x78, PHY:
> 0x45)
> [ 12.849404] r8169 0000:00:01.0 enp0s1: link down
> [ 12.849432] r8169 0000:00:01.0 enp0s1: link down
> [ 12.849474] IPv6: ADDRCONF(NETDEV_UP): enp0s1: link is not ready
> [ 12.892154] r8169 0000:00:02.0 enp0s2: unable to load firmware patch
> rtl_nic/rtl8168d-1.fw (-2)

Strange, the linux-firmware contains this file. Anyway this particular error
is about wired interface, not the wifi one.

> [ 12.905427] r8169 0000:00:02.0 enp0s2: link down
> [ 12.905467] IPv6: ADDRCONF(NETDEV_UP): enp0s2: link is not ready
> [ 12.913229] ath5k: phy0: ath5k_rx_skb_alloc: DMA mapping failed
> [ 12.913240] ath5k: phy0: can't start recv logic
> [ 13.357821] ath5k: phy0: ath5k_rx_skb_alloc: DMA mapping failed
> [ 13.357831] ath5k: phy0: can't start recv logic
> [ 13.363317] ath5k: phy0: ath5k_rx_skb_alloc: DMA mapping failed
> [ 13.363324] ath5k: phy0: can't start recv logic
> [ 13.369350] ath5k: phy0: ath5k_rx_skb_alloc: DMA mapping failed
> [ 13.369357] ath5k: phy0: can't start recv logic
> [ 13.374856] ath5k: phy0: ath5k_rx_skb_alloc: DMA mapping failed
> [ 13.374867] ath5k: phy0: can't start recv logic

This suggests mi one other possibility: perhaps you need to increase swiotlb
parameter to netvm kernel (amount of memory used for DMA). In dom0 console try
this:
qvm-prefs -s netvm kernelopts "iommu=soft swiotlb=16384"
(by default there is 8192)
Then reboot.

(...)

> 4. Check if linux-firmware package is installed (rpm -q linux-firmware). -
> YES
>
> [user@fedora-20-x64 ~]$ rpm -q linux-firmware
> linux-firmware-20140605-38.gita4f3bc03.fc20.noarch
>
> For wirting this mail I need to connect ethernet cable, disable firewallvm.
> The WiFI adapter sees the connection but wen I try to enter the pass, its
> always fail (And yes the pass is correct :D )
>
> Please could someone, point-me in the right direction please.
>
> Thanks in advance,
>
> BS
>
> Fernando.

[Jos den Bekker]

on 8-9-2014 23:55 Marek Marczykowski-Górecki wrote:
> Ok, so try to unload and load again tg3 module.

I forgot to tell that I restarted after, so tg3 module was reloaded. Still no internet connection.

> Generally, you should install that package in the template VM (fedora-20-x64) to have that change persistent.

Did that. Incidently, when I started Qubes just now, linux-firmware wasn't there anymore:

[user@netvm ~]$ rpm -q linux-firmware

package linux-firmware is not installed
[user@netvm ~]$

Then I cahnged to TemplateVM:

[user@fedora-20-x64 ~]$ rpm -q linux-firmware
package linux-firmware is not installed

I installed the package:

[user@fedora-20-x64 ~]$ sudo rpm -i /run/media/user/CRUZER/linux-firmware-20140605-38.gita4f3bc03.fc20.noarch.rpm
[user@fedora-20-x64 ~]$ rpm -q linux-firmware
linux-firmware-20140605-38.gita4f3bc03.fc20.noarch
[user@fedora-20-x64 ~]$

Then I restarted Qubes again:

[user@fedora-20-x64 ~]$ rpm -q linux-firmware
linux-firmware-20140605-38.gita4f3bc03.fc20.noarch
[user@fedora-20-x64 ~]$

So now it's there persistently (I hope). Still no internet connection. dmesg gave same error messages for tg3:

[    9.889495] tg3.c:v3.133 (Jul 29, 2013)
[    9.889743] tg3 0000:00:00.0: enabling device (0000 -> 0002)
[    9.890154] tg3 0000:00:00.0: Xen PCI mapped GSI17 to IRQ51
[    9.929392] libphy: tg3 mdio bus: probed
[    9.946123] tg3 0000:00:00.0: Could not obtain valid ethernet address, aborting
[    9.946317] tg3: probe of 0000:00:00.0 failed with error -22

Fernando Parodi said he could gain access via ethernet cable by disabling firewallVM. Haven't tried that. Should I? Would a change in firewall settings help? I'm a bit puzzled.

Thanks in advance for any help.

Jos

[Marek Marczykowski-Górecki]

Just to be sure - now your netvm have linux-firmware package (that installed
in the template), right?

> Fernando Parodi said he could gain access via ethernet cable by disabling
> firewallVM. Haven't tried that. Should I? Would a change in firewall
> settings help? I'm a bit puzzled.

Actually I have no idea what he meant by "disabling firewallVM"...

One another idea - check dom0 dmesg for messages from pciback driver (some
errors about config space access?).

[Fernando Parodi]

Actually I have no idea what he meant by "disabling firewallVM"...

On Tue, Sep 9, 2014 at 3:04 PM, Marek Marczykowski-Górecki <marm...@invisiblethingslab.com> wrote:

I Mean i shutdown the VM, and the change the netvm in the setting of the VM were i send the previous mail.

Actually I'M having 3 Problems:

1 - Wifi sees my connection but when i enter the <correct> pass always fails
2 - If i dont shutdown the FirewallVM I dont have any conn.

3-  For the wired conn: start working ok, but a few minutes after i have several micro-cuts.

When I leave my work, i will try increase the buffer as you mention early.

[Marek Marczykowski-Górecki]

On 09.09.2014 21:27, Fernando Parodi wrote:
> Actually I have no idea what he meant by "disabling firewallVM"...
>
> On Tue, Sep 9, 2014 at 3:04 PM, Marek Marczykowski-Górecki <
> marm...@invisiblethingslab.com> wrote:
>
> I Mean i shutdown the VM, and the change the netvm in the setting of the VM
> were i send the previous mail.
>
> Actually I'M having 3 Problems:
>
> 1 - Wifi sees my connection but when i enter the <correct> pass always fails

Anything interesting in logs (/var/log/messages in netvm)?

> 2 - If i dont shutdown the FirewallVM I dont have any conn.

Maybe you've set firewall settings to block all the traffic? Check the VM
settings, make sure you've also enabled DNS access.

[Jos den Bekker]

on 9-9-2014 20:04 Marek Marczykowski-Górecki wrote:

Just to be sure - now your netvm have linux-firmware package (that
installed in the template), right? 

Yep!

One another idea - check dom0 dmesg for messages from pciback driver
(some errors about config space access?). 

Here is what it says regarding pciback:

[ 35.375830] pciback 0000:04:00.0: Driver tried to write to a read-only
configuration space field at offset 0x52, size 2. This may be harmless,
but if you have problems with your device:
1) see permissive attribute in sysfs
2) report problems to the xen-devel mailing list along with details of
your device obtained from lspci.
...
[44.081163] pciback 0000:04:00.0: enabling device (0000 -> 0002)

Best regards and thanks so far!

Jos

[Marek Marczykowski-Górecki]

Ok, so you have instructions here what to do.

[Fernando Parodi]

This suggests mi one other possibility: perhaps you need to increase swiotlb
parameter to netvm kernel (amount of memory used for DMA). In dom0 console try
this:
qvm-prefs -s netvm kernelopts "iommu=soft swiotlb=16384"
(by default there is 8192)
Then reboot.

Thanks Marek, that command solve all my internet conn problems *Wifi, Wired and Firewallvm(

These mail I sended in my Qubes with the Wifi >D

Thanks  por the help.

[Jos den Bekker]

on 9-9-2014 23:06 Marek Marczykowski-Górecki wrote:

> Ok, so you have instructions here what to do.

Not really. ;-)  First, I don't know what change, if any, to make to the permissive attribute in sysfs, and I hate to be doing anything if I don't know what I'm doing. And second, I posted my query on the xen-devel mailing list, but that list is NOT for technical problems, it is ONLY for developers, so they won't answer me.

Any suggestions anybody?

Thanks,

Jos

[Jos den Bekker]

Let me update (or should I start another thread?).

First off, I don't think I gave full specifications for my ethernetcard. So, lspci renders:

Broadcom Corporation Netlink BCM57780 Gigabit Ethernet PCIe (rev 01)

I did the same thing as Fernando, ie in dom0 I issued the command:

qvm-prefs -s netvm kernelopts "iommu=soft swiotlb=16384"

Restarted. Made no difference.

Then I found this in the Xen Wiki (wiki.xen.org/wiki/Xen_PCI_Passthrough):

To enable permissive mode for a device using xl, you can enable it for all devices for a given domain in the /etc/xen/<domain> configuration file, like this:

pci_permissive=1

Or, you can add ",permissive=1" to the BDF of a particular device as it's passed through, either in the config file:

pci=['08:00.0,permissive=1']

or when hot-plugging it:

xl pci-attach 5 '08:00.0,permissive=1'


 Permissive mode for xm/xend

xend works a little differently than xl; rather than specifying a particular device for a particular domain, it has a global list of devices which it allowes to be set as "permissive" in /etc/xen/xend-pci-permissive.sxp.

First, find the hexadecimal vendor id for the device you want to pass through using lspci -nn:

# lspci -nn
08:04.0 Ethernet controller [0200]: Broadcom Corporation NetXtreme BCM5715 Gigabit Ethernet [14e4:1678] (rev a3)

So the hex vendor id is 14e4:1678. Now add that code to /etc/xen/xend-pci-permissive.sxp, in the section called unconstrained_dev_ids. The end result should look like this:

(unconstrained_dev_ids
    ('14e4:1678')
)

/etc/xen/<domain> doesn't exist, but /etc/xen/xend-pci-permissive.sxp does, so I assume I have xend. "lspci -nn" gave the hex vendor id "14e4:1692". I added that to /etc/xen/xend-pci-permissive.sxp exactly as described above, restarted, but again no internet connection.

I should add, perhaps, that the error message I get after startup says "internet connection has been (or was) shut down". Not something like "could not establish internet connection".

Again, any help is welcome.

Jos

[Marek Marczykowski-Górecki]

On 11.09.2014 12:06, Jos den Bekker wrote:
> Let me update (or should I start another thread?).
>
> First off, I don't think I gave full specifications for my ethernetcard.
> So, lspci renders:
>
> Broadcom Corporation Netlink BCM57780 Gigabit Ethernet PCIe (rev 01)
>
> I did the same thing as Fernando, ie in dom0 I issued the command:
>
> qvm-prefs -s netvm kernelopts "iommu=soft swiotlb=16384"
>
> Restarted. Made no difference.
>
> Then I found this in the Xen Wiki (wiki.xen.org/wiki/Xen_PCI_Passthrough):
>
> To enable permissive mode for a device using xl, you can enable it for all
> devices for a given domain in the /etc/xen/<domain> configuration file,
> like this:
>
> pci_permissive=1
>
> Or, you can add ",permissive=1" to the BDF of a particular device as it's
> passed through, either in the config file:
>
> pci=['08:00.0,permissive=1']

You can modify Xen config by copying autogenerated one
(/var/lib/qubes/servicevms/netvm/netvm.conf) and starting the netvm with
qvm-start --custom-config=/path/to/modified/config.

>
> or when hot-plugging it:
>
> xl pci-attach 5 '08:00.0,permissive=1'

This option could be even simpler, but IMO less reliable (also remember to
detach the device first).

Alternatively, you can enable permissive mode directly in sysfs:
echo BDF-OF-THE-DEVICE > /sys/bus/pci/drivers/pciback/permissive

[Jos den Bekker]

On 11 September 2014 12:41:08 UTC+2 Marek Marczykowski-Górecki wrote:

You can modify Xen config by copying autogenerated one
(/var/lib/qubes/servicevms/netvm/netvm.conf) and starting the netvm with
qvm-start --custom-config=/path/to/modified/config.

How should I modify netvm.conf? The command you specified is incorrect, the right syntax is "qvm-start [options] [vm]. So in this form it will say that no VM is specified. Should I stop netvm and restart it this way? (After I've modified the file, which I don't know how, because it says nothing about permissiveness. Sorry to be so dumb.)

Alternatively, you can enable permissive mode directly in sysfs:
echo BDF-OF-THE-DEVICE > /sys/bus/pci/drivers/pciback/permissive

Doesn't work: "permission denied". (Doesn't make any difference if I use "sudo" or "sudo -v -u root" + password.) I can load the file in a text editor (nano), but not save it: "permission denied".


Best regards, Jos

[Marek Marczykowski-Górecki]

On 11.09.2014 15:38, Jos den Bekker wrote:
>
>
> On 11 September 2014 12:41:08 UTC+2 Marek Marczykowski-Górecki wrote:
>
> You can modify Xen config by copying autogenerated one
>> (/var/lib/qubes/servicevms/netvm/netvm.conf) and starting the netvm with
>> qvm-start --custom-config=/path/to/modified/config.
>>
>
> How should I modify netvm.conf? The command you specified is incorrect, the
> right syntax is "qvm-start [options] [vm]. So in this form it will say that
> no VM is specified. Should I stop netvm and restart it this way? (After
> I've modified the file, which I don't know how, because it says nothing
> about permissiveness. Sorry to be so dumb.)

Yes, you need to copy netvm.conf (the original netvm.conf is overwritten at
each VM startup), then restart netvm using above command (yes, add "netvm" at
the end).

> Alternatively, you can enable permissive mode directly in sysfs:
>> echo BDF-OF-THE-DEVICE > /sys/bus/pci/drivers/pciback/permissive
>>
>
> Doesn't work: "permission denied". (Doesn't make any difference if I use
> "sudo" or "sudo -v -u root" + password.) I can load the file in a text
> editor (nano), but not save it: "permission denied".

Swich to root with "sudo -s", then enter the command.

[Jos den Bekker]

On 11 september 2014 15:44:57 UTC+2 Marek Marczykowski-Górecki wrote:

Yes, you need to copy netvm.conf (the original netvm.conf is overwritten at
each VM startup), then restart netvm using above command (yes, add "netvm" at
the end).

OK. Shut down netvm and all its dependencies, then restarted netvm with copied conf file. Made no difference. (Why should it?)
 

Swich to root with "sudo -s", then enter the command.

Error message saying: "echo: wrong parameter". Strange. As far as I can see, /sys/bus/pci/drivers/pciback/permissive is a file, so BDF-OF-THE-DEVICE should be written to it (or appended with >>, which I tried, but gave the same result). I loaded the file in nano, wrote BDF-OF-THE-DEVICE in it, then saves, but nano said: "couldn't save file: wrong parameter". Excuse me? I'm quite puzzled. But still nothing resembling a network connection.

Could there be something fundamentally wrong with my installation of Qubes?

Best regards, Jos

[Marek Marczykowski-Górecki]

Uhm... replace BDF-OF-THE-DEVICE with actuall BDF (for example 0000:00:19.0).

[Jos den Bekker]

On 11 september 2014 17:06:21 UTC+2 Marek Marczykowski-Górecki wrote:

Uhm... replace BDF-OF-THE-DEVICE with actuall BDF (for example 0000:00:19.0).

Finally worked. Don't know what went wrong yesterday, maybe I should have shut down netvm before making the change. Now I was connected and could finally install updates. But the change in the permissive file is gone at every boot. Is there a way to make the change in the file permanent?

Thanks!

[Marek Marczykowski-Górecki]

You can add that command to /etc/rc.local in dom0.

[q...@caltech.edu]

...you may edit in dom-0 as su

/etc/systemd/system/multi-user.target.wants/qubes-netvm.service

...and add in line 11 of the [Service] section

ExecStart=/bin/sh -c 'echo 0000:04:00.0 > /sys/bus/pci/drivers/pciback/permissive'
...followed by the NetVm Start command line

...reStart firewallvm/netvm ...done

[Marek Marczykowski-Górecki]

Better to add new ExecStartPre (with above command). But note that this file
will be overwritten during system update. To prevent this, you can create
simple new service. Create /etc/systemd/system/qubes-pre-netvm.service with:
-----
[Unit]
Description=Netvm fixup
Before=qubes-netvm.service

[Service]

ExecStart=/bin/sh -c 'echo 0000:04:00.0 > /sys/bus/pci/drivers/pciback/permissive'

Type=oneshot
RemainAfterExit=yes

[Install]
WantedBy=multi-user.target
-----
Then enable it with "systemctl enable qubes-pre-netvm.service".

[Jos den Bekker]

On 12 september 2014 22:30:32 UTC+2 Marek Marczykowski-Górecki wrote:

Better to add new ExecStartPre (with above command). But note that this file
will be overwritten during system update. To prevent this, you can create
simple new service. Create /etc/systemd/system/qubes-pre-netvm.service with:
-----
[Unit]
Description=Netvm fixup
Before=qubes-netvm.service

[Service]
ExecStart=/bin/sh -c 'echo 0000:04:00.0 > /sys/bus/pci/drivers/pciback/permissive'
Type=oneshot
RemainAfterExit=yes

[Install]
WantedBy=multi-user.target
-----
Then enable it with "systemctl enable qubes-pre-netvm.service".

That worked wonderfully well, Marek. Thanks very very much!! (Only the name of the netvm service has an underscore: qubes_netvm.service. :-))

Beste regards, Jos

[John Doe]

Could someone outline, in layman's terms (explicit commands, numbered steps), what the fix was? I'm having a hard time following what was done in this thread. I have determined that I have the same Ethernet controller.

Is this a driver issue? Why does Qubes require CLI magic just to get an internet connection out-of-box? It's a huge turn off, I've been trying to figure this out for 3 days.

[hast...@gmail.com]

Thank you,

today this post helped me a lot... This worked for me too. Thank you also for the "_" correction:-)

Steffen

[Drew White]

On Tuesday, 8 March 2016 06:35:07 UTC+11, Lance Meredith wrote:
> Could someone outline, in layman's terms (explicit commands, numbered steps), what the fix was? I'm having a hard time following what was done in this thread. I have determined that I have the same Ethernet controller.
>
> Is this a driver issue? Why does Qubes require CLI magic just to get an internet connection out-of-box? It's a huge turn off, I've been trying to figure this out for 3 days.

AGREED, but noone ever thinks about that, because the system is designed for end-users, not coders, and yet they don't make allowances for the end-users that aren't coders.

I was told that the solution was here, all I see is a bunch of stuff that doesn't make sense to me. and I'm a programmer.

[andreazi...@gmail.com]

Hello I know it's been 5 years since this thread started but apple hasn't changed and problems are still the same.

I have MB air 2017 with broadcom bcm4360, i've read a lot on how to solve this wifi problem but i got so many different info that i can't follow a single guide.

Honestly i'm not the best in using the terminal but i can go through it.

I'd appreciate if someone could write a step by step guide in order to get this  wifi working.

Thank you all for yoru help